Power System State Recovery using Local and Global Smoothness of its Graph Signals

Recovering the state of unobservable power system components due to cyber attacks or limited meter availability is a crucial problem to address to enable efficient monitoring and operation of power systems. The graph signal processing (GSP) framework provides new opportunities to improve power system data analysis by capturing the topological information of the system. In this paper, the recovery of the unobservable states in power systems is formulated as a graph signal reconstruction problem in a GSP framework. Specifically, a novel reconstruction technique based on the statistics of the local smoothness of the graph signals along with the global smoothness of the graph signals is casted into an optimization framework. In contrast to many graph signal reconstruction techniques, which assume band-limited signals to be recovered, the proposed technique is applicable to general graph signals irrespective of their bandwidth. The performance evaluation of the proposed method using simulated graph signals for the IEEE 118 bus system show promising reconstruction accuracy.